Building a good model-router is technically challenging. It requires understanding and measuring model performance (both through qualitative metrics and human preferences), developing more effective routing systems (better data, new model architectures, query-rewriting, and fundamental research into how models work), and productionizing the tools we build (reducing latency, improving model costs, ML Ops). We're looking for someone to take on those challenges with Martian by building and leading our ML Eng org from scratch. Full details can be found here.